2 /* $KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $ */
5 * SPDX-License-Identifier: BSD-3-Clause
7 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the project nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * IPsec controller part.
40 #include "opt_inet6.h"
41 #include "opt_ipsec.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/malloc.h>
47 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/errno.h>
53 #include <sys/hhook.h>
55 #include <sys/kernel.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
61 #include <net/if_enc.h>
62 #include <net/if_var.h>
65 #include <netinet/in.h>
66 #include <netinet/in_systm.h>
67 #include <netinet/ip.h>
68 #include <netinet/ip_var.h>
69 #include <netinet/in_var.h>
70 #include <netinet/udp.h>
71 #include <netinet/udp_var.h>
72 #include <netinet/tcp.h>
73 #include <netinet/udp.h>
75 #include <netinet/ip6.h>
77 #include <netinet6/ip6_var.h>
79 #include <netinet/in_pcb.h>
81 #include <netinet/icmp6.h>
84 #include <sys/types.h>
85 #include <netipsec/ipsec.h>
87 #include <netipsec/ipsec6.h>
89 #include <netipsec/ah_var.h>
90 #include <netipsec/esp_var.h>
91 #include <netipsec/ipcomp.h> /*XXX*/
92 #include <netipsec/ipcomp_var.h>
93 #include <netipsec/ipsec_support.h>
95 #include <netipsec/key.h>
96 #include <netipsec/keydb.h>
97 #include <netipsec/key_debug.h>
99 #include <netipsec/xform.h>
101 #include <machine/in_cksum.h>
103 #include <opencrypto/cryptodev.h>
105 /* NB: name changed so netstat doesn't use it. */
106 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec4stat);
107 VNET_PCPUSTAT_SYSINIT(ipsec4stat);
110 VNET_PCPUSTAT_SYSUNINIT(ipsec4stat);
113 /* DF bit on encap. 0: clear 1: set 2: copy */
114 VNET_DEFINE(int, ip4_ipsec_dfbit) = 0;
115 VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE;
116 VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE;
117 VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE;
118 VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE;
119 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
120 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
122 VNET_DEFINE_STATIC(int, ip4_filtertunnel) = 0;
123 #define V_ip4_filtertunnel VNET(ip4_filtertunnel)
124 VNET_DEFINE_STATIC(int, check_policy_history) = 0;
125 #define V_check_policy_history VNET(check_policy_history)
126 VNET_DEFINE_STATIC(struct secpolicy *, def_policy) = NULL;
127 #define V_def_policy VNET(def_policy)
129 sysctl_def_policy(SYSCTL_HANDLER_ARGS)
133 value = V_def_policy->policy;
134 error = sysctl_handle_int(oidp, &value, 0, req);
136 if (value != IPSEC_POLICY_DISCARD &&
137 value != IPSEC_POLICY_NONE)
139 V_def_policy->policy = value;
145 * Crypto support requirements:
147 * 1 require hardware support
148 * -1 require software support
151 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
154 * Use asynchronous mode to parallelize crypto jobs:
159 VNET_DEFINE(int, async_crypto) = 0;
162 * TCP/UDP checksum handling policy for transport mode NAT-T (RFC3948)
164 * 0 - auto: incrementally recompute, when checksum delta is known;
165 * if checksum delta isn't known, reset checksum to zero for UDP,
166 * and mark csum_flags as valid for TCP.
167 * 1 - fully recompute TCP/UDP checksum.
169 VNET_DEFINE(int, natt_cksum_policy) = 0;
171 FEATURE(ipsec, "Internet Protocol Security (IPsec)");
172 FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')");
174 SYSCTL_DECL(_net_inet_ipsec);
177 SYSCTL_PROC(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
178 CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
179 0, 0, sysctl_def_policy, "I",
180 "IPsec default policy.");
181 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
182 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
183 "Default ESP transport mode level");
184 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
185 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
186 "Default ESP tunnel mode level.");
187 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
188 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
189 "AH transfer mode default level.");
190 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
191 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
192 "AH tunnel mode default level.");
193 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
194 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
195 "If set, clear type-of-service field when doing AH computation.");
196 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
197 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
198 "Do not fragment bit on encap.");
199 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
200 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
201 "Explicit Congestion Notification handling.");
202 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
203 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
204 "Crypto driver selection.");
205 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, async_crypto,
206 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(async_crypto), 0,
207 "Use asynchronous mode to parallelize crypto jobs.");
208 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, check_policy_history,
209 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(check_policy_history), 0,
210 "Use strict check of inbound packets to security policy compliance.");
211 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, natt_cksum_policy,
212 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(natt_cksum_policy), 0,
213 "Method to fix TCP/UDP checksum for transport mode IPsec after NAT.");
214 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, filtertunnel,
215 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_filtertunnel), 0,
216 "If set, filter packets from an IPsec tunnel.");
217 SYSCTL_VNET_PCPUSTAT(_net_inet_ipsec, OID_AUTO, ipsecstats, struct ipsecstat,
218 ipsec4stat, "IPsec IPv4 statistics.");
220 struct timeval ipsec_warn_interval = { .tv_sec = 1, .tv_usec = 0 };
221 SYSCTL_TIMEVAL_SEC(_net_inet_ipsec, OID_AUTO, crypto_warn_interval, CTLFLAG_RW,
222 &ipsec_warn_interval,
223 "Delay in seconds between warnings of deprecated IPsec crypto algorithms.");
227 * When set to 1, IPsec will send packets with the same sequence number.
228 * This allows to verify if the other side has proper replay attacks detection.
230 VNET_DEFINE(int, ipsec_replay) = 0;
231 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_replay,
232 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
233 "Emulate replay attack");
235 * When set 1, IPsec will send packets with corrupted HMAC.
236 * This allows to verify if the other side properly detects modified packets.
238 VNET_DEFINE(int, ipsec_integrity) = 0;
239 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
240 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
241 "Emulate man-in-the-middle attack");
245 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec6stat);
246 VNET_PCPUSTAT_SYSINIT(ipsec6stat);
249 VNET_PCPUSTAT_SYSUNINIT(ipsec6stat);
252 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
253 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
254 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
255 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
256 VNET_DEFINE(int, ip6_ipsec_ecn) = 0; /* ECN ignore(-1)/forbidden(0)/allowed(1) */
258 VNET_DEFINE_STATIC(int, ip6_filtertunnel) = 0;
259 #define V_ip6_filtertunnel VNET(ip6_filtertunnel)
261 SYSCTL_DECL(_net_inet6_ipsec6);
263 /* net.inet6.ipsec6 */
264 SYSCTL_PROC(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy,
265 CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
266 0, 0, sysctl_def_policy, "I",
267 "IPsec default policy.");
268 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
269 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev), 0,
270 "Default ESP transport mode level.");
271 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
272 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev), 0,
273 "Default ESP tunnel mode level.");
274 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
275 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev), 0,
276 "AH transfer mode default level.");
277 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
278 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev), 0,
279 "AH tunnel mode default level.");
280 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN, ecn,
281 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn), 0,
282 "Explicit Congestion Notification handling.");
283 SYSCTL_INT(_net_inet6_ipsec6, OID_AUTO, filtertunnel,
284 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_filtertunnel), 0,
285 "If set, filter packets from an IPsec tunnel.");
286 SYSCTL_VNET_PCPUSTAT(_net_inet6_ipsec6, IPSECCTL_STATS, ipsecstats,
287 struct ipsecstat, ipsec6stat, "IPsec IPv6 statistics.");
290 static int ipsec_in_reject(struct secpolicy *, struct inpcb *,
291 const struct mbuf *);
294 static void ipsec4_get_ulp(const struct mbuf *, struct secpolicyindex *, int);
295 static void ipsec4_setspidx_ipaddr(const struct mbuf *,
296 struct secpolicyindex *);
299 static void ipsec6_get_ulp(const struct mbuf *m, struct secpolicyindex *, int);
300 static void ipsec6_setspidx_ipaddr(const struct mbuf *,
301 struct secpolicyindex *);
305 * Return a held reference to the default SP.
307 static struct secpolicy *
308 key_allocsp_default(void)
311 key_addref(V_def_policy);
312 return (V_def_policy);
316 ipsec_invalidate_cache(struct inpcb *inp, u_int dir)
318 struct secpolicy *sp;
320 INP_WLOCK_ASSERT(inp);
321 if (dir == IPSEC_DIR_OUTBOUND) {
322 if (inp->inp_sp->flags & INP_INBOUND_POLICY)
324 sp = inp->inp_sp->sp_in;
325 inp->inp_sp->sp_in = NULL;
327 if (inp->inp_sp->flags & INP_OUTBOUND_POLICY)
329 sp = inp->inp_sp->sp_out;
330 inp->inp_sp->sp_out = NULL;
333 key_freesp(&sp); /* release extra reference */
337 ipsec_cachepolicy(struct inpcb *inp, struct secpolicy *sp, u_int dir)
342 INP_LOCK_ASSERT(inp);
344 if (dir == IPSEC_DIR_OUTBOUND) {
345 /* Do we have configured PCB policy? */
346 if (inp->inp_sp->flags & INP_OUTBOUND_POLICY)
348 /* Another thread has already set cached policy */
349 if (inp->inp_sp->sp_out != NULL)
352 * Do not cache OUTBOUND policy if PCB isn't connected,
353 * i.e. foreign address is INADDR_ANY/UNSPECIFIED.
356 if ((inp->inp_vflag & INP_IPV4) != 0 &&
357 inp->inp_faddr.s_addr == INADDR_ANY)
361 if ((inp->inp_vflag & INP_IPV6) != 0 &&
362 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
366 /* Do we have configured PCB policy? */
367 if (inp->inp_sp->flags & INP_INBOUND_POLICY)
369 /* Another thread has already set cached policy */
370 if (inp->inp_sp->sp_in != NULL)
373 * Do not cache INBOUND policy for listen socket,
374 * that is bound to INADDR_ANY/UNSPECIFIED address.
377 if ((inp->inp_vflag & INP_IPV4) != 0 &&
378 inp->inp_faddr.s_addr == INADDR_ANY)
382 if ((inp->inp_vflag & INP_IPV6) != 0 &&
383 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
388 if (!INP_WLOCKED(inp)) {
389 if ((downgrade = INP_TRY_UPGRADE(inp)) == 0)
392 if (dir == IPSEC_DIR_OUTBOUND)
393 inp->inp_sp->sp_out = sp;
395 inp->inp_sp->sp_in = sp;
397 * SP is already referenced by the lookup code.
398 * We take extra reference here to avoid race in the
399 * ipsec_getpcbpolicy() function - SP will not be freed in the
400 * time between we take SP pointer from the cache and key_addref()
404 genid = key_getspgen();
405 if (genid != inp->inp_sp->genid) {
406 ipsec_invalidate_cache(inp, dir);
407 inp->inp_sp->genid = genid;
410 printf("%s: PCB(%p): cached %s SP(%p)\n",
411 __func__, inp, dir == IPSEC_DIR_OUTBOUND ? "OUTBOUND":
417 static struct secpolicy *
418 ipsec_checkpolicy(struct secpolicy *sp, struct inpcb *inp, int *error)
421 /* Save found OUTBOUND policy into PCB SP cache. */
422 if (inp != NULL && inp->inp_sp != NULL && inp->inp_sp->sp_out == NULL)
423 ipsec_cachepolicy(inp, sp, IPSEC_DIR_OUTBOUND);
425 switch (sp->policy) {
427 printf("%s: invalid policy %u\n", __func__, sp->policy);
429 case IPSEC_POLICY_DISCARD:
430 *error = -EINVAL; /* Packet is discarded by caller. */
432 case IPSEC_POLICY_BYPASS:
433 case IPSEC_POLICY_NONE:
435 sp = NULL; /* NB: force NULL result. */
437 case IPSEC_POLICY_IPSEC:
438 /* XXXAE: handle LARVAL SP */
442 printf("%s: get SP(%p), error %d\n", __func__, sp, *error));
446 static struct secpolicy *
447 ipsec_getpcbpolicy(struct inpcb *inp, u_int dir)
449 struct secpolicy *sp;
450 int flags, downgrade;
452 if (inp == NULL || inp->inp_sp == NULL)
455 INP_LOCK_ASSERT(inp);
457 flags = inp->inp_sp->flags;
458 if (dir == IPSEC_DIR_OUTBOUND) {
459 sp = inp->inp_sp->sp_out;
460 flags &= INP_OUTBOUND_POLICY;
462 sp = inp->inp_sp->sp_in;
463 flags &= INP_INBOUND_POLICY;
466 * Check flags. If we have PCB SP, just return it.
467 * Otherwise we need to check that cached SP entry isn't stale.
472 if (inp->inp_sp->genid != key_getspgen()) {
473 /* Invalidate the cache. */
475 if (!INP_WLOCKED(inp)) {
476 if ((downgrade = INP_TRY_UPGRADE(inp)) == 0)
479 ipsec_invalidate_cache(inp, IPSEC_DIR_OUTBOUND);
480 ipsec_invalidate_cache(inp, IPSEC_DIR_INBOUND);
486 printf("%s: PCB(%p): cache hit SP(%p)\n",
488 /* Return referenced cached policy */
496 ipsec4_get_ulp(const struct mbuf *m, struct secpolicyindex *spidx,
503 IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),
504 ("packet too short"));
506 if (m->m_len >= sizeof (struct ip)) {
507 const struct ip *ip = mtod(m, const struct ip *);
508 if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
510 off = ip->ip_hl << 2;
515 m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
516 if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
522 while (off < m->m_pkthdr.len) {
529 spidx->ul_proto = nxt;
532 if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
534 m_copydata(m, off, sizeof (th), (caddr_t) &th);
535 spidx->src.sin.sin_port = th.th_sport;
536 spidx->dst.sin.sin_port = th.th_dport;
539 spidx->ul_proto = nxt;
542 if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
544 m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
545 spidx->src.sin.sin_port = uh.uh_sport;
546 spidx->dst.sin.sin_port = uh.uh_dport;
549 if (off + sizeof(ip6e) > m->m_pkthdr.len)
551 /* XXX Sigh, this works but is totally bogus. */
552 m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
553 off += (ip6e.ip6e_len + 2) << 2;
558 /* XXX Intermediate headers??? */
559 spidx->ul_proto = nxt;
564 spidx->ul_proto = IPSEC_ULPROTO_ANY;
566 spidx->src.sin.sin_port = IPSEC_PORT_ANY;
567 spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
569 printf("%s: ", __func__); kdebug_secpolicyindex(spidx, NULL));
573 ipsec4_setspidx_ipaddr(const struct mbuf *m, struct secpolicyindex *spidx)
576 ipsec4_setsockaddrs(m, &spidx->src, &spidx->dst);
577 spidx->prefs = sizeof(struct in_addr) << 3;
578 spidx->prefd = sizeof(struct in_addr) << 3;
581 static struct secpolicy *
582 ipsec4_getpolicy(const struct mbuf *m, struct inpcb *inp, u_int dir,
585 struct secpolicyindex spidx;
586 struct secpolicy *sp;
588 sp = ipsec_getpcbpolicy(inp, dir);
589 if (sp == NULL && key_havesp(dir)) {
590 /* Make an index to look for a policy. */
591 ipsec4_setspidx_ipaddr(m, &spidx);
592 ipsec4_get_ulp(m, &spidx, needport);
594 sp = key_allocsp(&spidx, dir);
596 if (sp == NULL) /* No SP found, use system default. */
597 sp = key_allocsp_default();
602 * Check security policy for *OUTBOUND* IPv4 packet.
605 ipsec4_checkpolicy(const struct mbuf *m, struct inpcb *inp, int *error,
608 struct secpolicy *sp;
611 sp = ipsec4_getpolicy(m, inp, IPSEC_DIR_OUTBOUND, needport);
613 sp = ipsec_checkpolicy(sp, inp, error);
616 case 0: /* No IPsec required: BYPASS or NONE */
619 IPSECSTAT_INC(ips_out_polvio);
622 IPSECSTAT_INC(ips_out_inval);
626 printf("%s: using SP(%p), error %d\n", __func__, sp, *error));
628 KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
633 * Check IPv4 packet against *INBOUND* security policy.
634 * This function is called from tcp_input(), udp_input(),
635 * rip_input() and sctp_input().
638 ipsec4_in_reject(const struct mbuf *m, struct inpcb *inp)
640 struct secpolicy *sp;
643 sp = ipsec4_getpolicy(m, inp, IPSEC_DIR_INBOUND, 0);
644 result = ipsec_in_reject(sp, inp, m);
647 IPSECSTAT_INC(ips_in_polvio);
652 * IPSEC_CAP() method implementation for IPv4.
655 ipsec4_capability(struct mbuf *m, u_int cap)
659 case IPSEC_CAP_BYPASS_FILTER:
661 * Bypass packet filtering for packets previously handled
664 if (!V_ip4_filtertunnel &&
665 m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
668 case IPSEC_CAP_OPERABLE:
669 /* Do we have active security policies? */
670 if (key_havesp(IPSEC_DIR_INBOUND) != 0 ||
671 key_havesp(IPSEC_DIR_OUTBOUND) != 0)
682 ipsec6_get_ulp(const struct mbuf *m, struct secpolicyindex *spidx,
690 IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip6_hdr),
691 ("packet too short"));
694 spidx->ul_proto = IPSEC_ULPROTO_ANY;
695 spidx->src.sin6.sin6_port = IPSEC_PORT_ANY;
696 spidx->dst.sin6.sin6_port = IPSEC_PORT_ANY;
699 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
700 if (off < 0 || m->m_pkthdr.len < off)
705 spidx->ul_proto = nxt;
708 if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
710 m_copydata(m, off, sizeof(th), (caddr_t)&th);
711 spidx->src.sin6.sin6_port = th.th_sport;
712 spidx->dst.sin6.sin6_port = th.th_dport;
715 spidx->ul_proto = nxt;
718 if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
720 m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
721 spidx->src.sin6.sin6_port = uh.uh_sport;
722 spidx->dst.sin6.sin6_port = uh.uh_dport;
725 spidx->ul_proto = nxt;
726 if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
728 m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
729 spidx->src.sin6.sin6_port = htons((uint16_t)ih.icmp6_type);
730 spidx->dst.sin6.sin6_port = htons((uint16_t)ih.icmp6_code);
733 /* XXX Intermediate headers??? */
734 spidx->ul_proto = nxt;
738 printf("%s: ", __func__); kdebug_secpolicyindex(spidx, NULL));
742 ipsec6_setspidx_ipaddr(const struct mbuf *m, struct secpolicyindex *spidx)
745 ipsec6_setsockaddrs(m, &spidx->src, &spidx->dst);
746 spidx->prefs = sizeof(struct in6_addr) << 3;
747 spidx->prefd = sizeof(struct in6_addr) << 3;
750 static struct secpolicy *
751 ipsec6_getpolicy(const struct mbuf *m, struct inpcb *inp, u_int dir,
754 struct secpolicyindex spidx;
755 struct secpolicy *sp;
757 sp = ipsec_getpcbpolicy(inp, dir);
758 if (sp == NULL && key_havesp(dir)) {
759 /* Make an index to look for a policy. */
760 ipsec6_setspidx_ipaddr(m, &spidx);
761 ipsec6_get_ulp(m, &spidx, needport);
763 sp = key_allocsp(&spidx, dir);
765 if (sp == NULL) /* No SP found, use system default. */
766 sp = key_allocsp_default();
771 * Check security policy for *OUTBOUND* IPv6 packet.
774 ipsec6_checkpolicy(const struct mbuf *m, struct inpcb *inp, int *error,
777 struct secpolicy *sp;
780 sp = ipsec6_getpolicy(m, inp, IPSEC_DIR_OUTBOUND, needport);
782 sp = ipsec_checkpolicy(sp, inp, error);
785 case 0: /* No IPsec required: BYPASS or NONE */
788 IPSEC6STAT_INC(ips_out_polvio);
791 IPSEC6STAT_INC(ips_out_inval);
795 printf("%s: using SP(%p), error %d\n", __func__, sp, *error));
797 KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
802 * Check IPv6 packet against inbound security policy.
803 * This function is called from tcp6_input(), udp6_input(),
804 * rip6_input() and sctp_input().
807 ipsec6_in_reject(const struct mbuf *m, struct inpcb *inp)
809 struct secpolicy *sp;
812 sp = ipsec6_getpolicy(m, inp, IPSEC_DIR_INBOUND, 0);
813 result = ipsec_in_reject(sp, inp, m);
816 IPSEC6STAT_INC(ips_in_polvio);
821 * IPSEC_CAP() method implementation for IPv6.
824 ipsec6_capability(struct mbuf *m, u_int cap)
828 case IPSEC_CAP_BYPASS_FILTER:
830 * Bypass packet filtering for packets previously handled
833 if (!V_ip6_filtertunnel &&
834 m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
837 case IPSEC_CAP_OPERABLE:
838 /* Do we have active security policies? */
839 if (key_havesp(IPSEC_DIR_INBOUND) != 0 ||
840 key_havesp(IPSEC_DIR_OUTBOUND) != 0)
849 ipsec_run_hhooks(struct ipsec_ctx_data *ctx, int type)
856 idx = HHOOK_IPSEC_INET;
861 idx = HHOOK_IPSEC_INET6;
865 return (EPFNOSUPPORT);
867 if (type == HHOOK_TYPE_IPSEC_IN)
868 HHOOKS_RUN_IF(V_ipsec_hhh_in[idx], ctx, NULL);
870 HHOOKS_RUN_IF(V_ipsec_hhh_out[idx], ctx, NULL);
871 if (*ctx->mp == NULL)
877 * Return current level.
878 * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
881 ipsec_get_reqlevel(struct secpolicy *sp, u_int idx)
883 struct ipsecrequest *isr;
884 u_int esp_trans_deflev, esp_net_deflev;
885 u_int ah_trans_deflev, ah_net_deflev;
888 IPSEC_ASSERT(idx < sp->tcount, ("Wrong IPsec request index %d", idx));
889 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
890 #define IPSEC_CHECK_DEFAULT(lev) \
891 (((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE && \
892 (lev) != IPSEC_LEVEL_UNIQUE) \
894 log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
895 (lev), IPSEC_LEVEL_REQUIRE) : 0), \
896 (lev) = IPSEC_LEVEL_REQUIRE, (lev) : (lev))
899 * IPsec VTI uses unique security policy with fake spidx filled
900 * with zeroes. Just return IPSEC_LEVEL_REQUIRE instead of doing
901 * full level lookup for such policies.
903 if (sp->state == IPSEC_SPSTATE_IFNET) {
904 IPSEC_ASSERT(sp->req[idx]->level == IPSEC_LEVEL_UNIQUE,
905 ("Wrong IPsec request level %d", sp->req[idx]->level));
906 return (IPSEC_LEVEL_REQUIRE);
909 /* Set default level. */
910 switch (sp->spidx.src.sa.sa_family) {
913 esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
914 esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
915 ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
916 ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
921 esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
922 esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
923 ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
924 ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
928 panic("%s: unknown af %u",
929 __func__, sp->spidx.src.sa.sa_family);
932 #undef IPSEC_CHECK_DEFAULT
936 switch (isr->level) {
937 case IPSEC_LEVEL_DEFAULT:
938 switch (isr->saidx.proto) {
940 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
941 level = esp_net_deflev;
943 level = esp_trans_deflev;
946 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
947 level = ah_net_deflev;
949 level = ah_trans_deflev;
953 * We don't really care, as IPcomp document says that
954 * we shouldn't compress small packets.
956 level = IPSEC_LEVEL_USE;
959 panic("%s: Illegal protocol defined %u\n", __func__,
964 case IPSEC_LEVEL_USE:
965 case IPSEC_LEVEL_REQUIRE:
968 case IPSEC_LEVEL_UNIQUE:
969 level = IPSEC_LEVEL_REQUIRE;
973 panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
980 ipsec_check_history(const struct mbuf *m, struct secpolicy *sp, u_int idx)
982 struct xform_history *xh;
986 while ((mtag = m_tag_find(__DECONST(struct mbuf *, m),
987 PACKET_TAG_IPSEC_IN_DONE, mtag)) != NULL) {
988 xh = (struct xform_history *)(mtag + 1);
990 char buf[IPSEC_ADDRSTRLEN];
991 printf("%s: mode %s proto %u dst %s\n", __func__,
992 kdebug_secasindex_mode(xh->mode), xh->proto,
993 ipsec_address(&xh->dst, buf, sizeof(buf))));
994 if (xh->proto != sp->req[idx]->saidx.proto)
996 /* If SA had IPSEC_MODE_ANY, consider this as match. */
997 if (xh->mode != sp->req[idx]->saidx.mode &&
998 xh->mode != IPSEC_MODE_ANY)
1001 * For transport mode IPsec request doesn't contain
1002 * addresses. We need to use address from spidx.
1004 if (sp->req[idx]->saidx.mode == IPSEC_MODE_TRANSPORT) {
1005 if (key_sockaddrcmp_withmask(&xh->dst.sa,
1006 &sp->spidx.dst.sa, sp->spidx.prefd) != 0)
1009 if (key_sockaddrcmp(&xh->dst.sa,
1010 &sp->req[idx]->saidx.dst.sa, 0) != 0)
1013 return (0); /* matched */
1019 * Check security policy requirements against the actual
1020 * packet contents. Return one if the packet should be
1021 * reject as "invalid"; otherwiser return zero to have the
1022 * packet treated as "valid".
1029 ipsec_in_reject(struct secpolicy *sp, struct inpcb *inp, const struct mbuf *m)
1034 printf("%s: PCB(%p): using SP(%p)\n", __func__, inp, sp));
1035 KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1037 if (inp != NULL && inp->inp_sp != NULL && inp->inp_sp->sp_in == NULL)
1038 ipsec_cachepolicy(inp, sp, IPSEC_DIR_INBOUND);
1041 switch (sp->policy) {
1042 case IPSEC_POLICY_DISCARD:
1044 case IPSEC_POLICY_BYPASS:
1045 case IPSEC_POLICY_NONE:
1049 IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1050 ("invalid policy %u", sp->policy));
1053 * ipsec[46]_common_input_cb after each transform adds
1054 * PACKET_TAG_IPSEC_IN_DONE mbuf tag. It contains SPI, proto, mode
1055 * and destination address from saidx. We can compare info from
1056 * these tags with requirements in SP.
1058 for (i = 0; i < sp->tcount; i++) {
1060 * Do not check IPcomp, since IPcomp document
1061 * says that we shouldn't compress small packets.
1062 * IPComp policy should always be treated as being
1065 if (sp->req[i]->saidx.proto == IPPROTO_IPCOMP ||
1066 ipsec_get_reqlevel(sp, i) != IPSEC_LEVEL_REQUIRE)
1068 if (V_check_policy_history != 0 &&
1069 ipsec_check_history(m, sp, i) != 0)
1071 else switch (sp->req[i]->saidx.proto) {
1073 if ((m->m_flags & M_DECRYPTED) == 0) {
1075 printf("%s: ESP m_flags:%x\n", __func__,
1081 if ((m->m_flags & M_AUTHIPHDR) == 0) {
1083 printf("%s: AH m_flags:%x\n", __func__,
1090 return (0); /* Valid. */
1094 * Compute the byte size to be occupied by IPsec header.
1095 * In case it is tunnelled, it includes the size of outer IP header.
1098 ipsec_hdrsiz_internal(struct secpolicy *sp)
1103 KEYDBG(IPSEC_STAMP, printf("%s: using SP(%p)\n", __func__, sp));
1104 KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1106 switch (sp->policy) {
1107 case IPSEC_POLICY_DISCARD:
1108 case IPSEC_POLICY_BYPASS:
1109 case IPSEC_POLICY_NONE:
1113 IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1114 ("invalid policy %u", sp->policy));
1117 * XXX: for each transform we need to lookup suitable SA
1118 * and use info from SA to calculate headers size.
1119 * XXX: for NAT-T we need to cosider UDP header size.
1122 for (i = 0; i < sp->tcount; i++) {
1123 switch (sp->req[i]->saidx.proto) {
1125 size += esp_hdrsiz(NULL);
1128 size += ah_hdrsiz(NULL);
1130 case IPPROTO_IPCOMP:
1131 size += sizeof(struct ipcomp);
1135 if (sp->req[i]->saidx.mode == IPSEC_MODE_TUNNEL) {
1136 switch (sp->req[i]->saidx.dst.sa.sa_family) {
1139 size += sizeof(struct ip);
1144 size += sizeof(struct ip6_hdr);
1148 ipseclog((LOG_ERR, "%s: unknown AF %d in "
1149 "IPsec tunnel SA\n", __func__,
1150 sp->req[i]->saidx.dst.sa.sa_family));
1159 * Compute ESP/AH header size for protocols with PCB, including
1160 * outer IP header. Currently only tcp_output() uses it.
1163 ipsec_hdrsiz_inpcb(struct inpcb *inp)
1165 struct secpolicyindex spidx;
1166 struct secpolicy *sp;
1169 sp = ipsec_getpcbpolicy(inp, IPSEC_DIR_OUTBOUND);
1170 if (sp == NULL && key_havesp(IPSEC_DIR_OUTBOUND)) {
1171 ipsec_setspidx_inpcb(inp, &spidx, IPSEC_DIR_OUTBOUND);
1172 sp = key_allocsp(&spidx, IPSEC_DIR_OUTBOUND);
1175 sp = key_allocsp_default();
1176 sz = ipsec_hdrsiz_internal(sp);
1182 * Check the variable replay window.
1183 * ipsec_chkreplay() performs replay check before ICV verification.
1184 * ipsec_updatereplay() updates replay bitmap. This must be called after
1185 * ICV verification (it also performs replay check, which is usually done
1187 * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1189 * Based on RFC 6479. Blocks are 32 bits unsigned integers
1192 #define IPSEC_BITMAP_INDEX_MASK(w) (w - 1)
1193 #define IPSEC_REDUNDANT_BIT_SHIFTS 5
1194 #define IPSEC_REDUNDANT_BITS (1 << IPSEC_REDUNDANT_BIT_SHIFTS)
1195 #define IPSEC_BITMAP_LOC_MASK (IPSEC_REDUNDANT_BITS - 1)
1198 ipsec_chkreplay(uint32_t seq, struct secasvar *sav)
1200 const struct secreplay *replay;
1201 uint32_t wsizeb; /* Constant: window size. */
1202 int index, bit_location;
1204 IPSEC_ASSERT(sav != NULL, ("Null SA"));
1205 IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1207 replay = sav->replay;
1209 /* No need to check replay if disabled. */
1210 if (replay->wsize == 0)
1214 wsizeb = replay->wsize << 3;
1216 /* Sequence number of 0 is invalid. */
1220 /* First time is always okay. */
1221 if (replay->count == 0)
1224 /* Larger sequences are okay. */
1225 if (seq > replay->lastseq)
1228 /* Over range to check, i.e. too old or wrapped. */
1229 if (replay->lastseq - seq >= wsizeb)
1232 /* The sequence is inside the sliding window
1233 * now check the bit in the bitmap
1234 * bit location only depends on the sequence number
1236 bit_location = seq & IPSEC_BITMAP_LOC_MASK;
1237 index = (seq >> IPSEC_REDUNDANT_BIT_SHIFTS)
1238 & IPSEC_BITMAP_INDEX_MASK(replay->bitmap_size);
1240 /* This packet already seen? */
1241 if ((replay->bitmap)[index] & (1 << bit_location))
1247 * Check replay counter whether to update or not.
1252 ipsec_updatereplay(uint32_t seq, struct secasvar *sav)
1255 struct secreplay *replay;
1256 uint32_t wsizeb; /* Constant: window size. */
1257 int diff, index, bit_location;
1259 IPSEC_ASSERT(sav != NULL, ("Null SA"));
1260 IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1262 replay = sav->replay;
1264 if (replay->wsize == 0)
1265 goto ok; /* No need to check replay. */
1268 wsizeb = replay->wsize << 3;
1270 /* Sequence number of 0 is invalid. */
1274 /* The packet is too old, no need to update */
1275 if (wsizeb + seq < replay->lastseq)
1278 /* Now update the bit */
1279 index = (seq >> IPSEC_REDUNDANT_BIT_SHIFTS);
1281 /* First check if the sequence number is in the range */
1282 if (seq > replay->lastseq) {
1284 int index_cur = replay->lastseq >> IPSEC_REDUNDANT_BIT_SHIFTS;
1286 diff = index - index_cur;
1287 if (diff > replay->bitmap_size) {
1288 /* something unusual in this case */
1289 diff = replay->bitmap_size;
1292 for (id = 0; id < diff; ++id) {
1293 replay->bitmap[(id + index_cur + 1)
1294 & IPSEC_BITMAP_INDEX_MASK(replay->bitmap_size)] = 0;
1297 replay->lastseq = seq;
1300 index &= IPSEC_BITMAP_INDEX_MASK(replay->bitmap_size);
1301 bit_location = seq & IPSEC_BITMAP_LOC_MASK;
1303 /* this packet has already been received */
1304 if (replay->bitmap[index] & (1 << bit_location))
1307 replay->bitmap[index] |= (1 << bit_location);
1310 if (replay->count == ~0) {
1312 /* Set overflow flag. */
1315 /* Don't increment, no more packets accepted. */
1316 if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0) {
1317 if (sav->sah->saidx.proto == IPPROTO_AH)
1318 AHSTAT_INC(ahs_wrap);
1319 else if (sav->sah->saidx.proto == IPPROTO_ESP)
1320 ESPSTAT_INC(esps_wrap);
1324 ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1325 __func__, replay->overflow,
1326 ipsec_sa2str(sav, buf, sizeof(buf))));
1334 ipsec_updateid(struct secasvar *sav, crypto_session_t *new,
1335 crypto_session_t *old)
1337 crypto_session_t tmp;
1340 * tdb_cryptoid is initialized by xform_init().
1341 * Then it can be changed only when some crypto error occurred or
1342 * when SA is deleted. We stored used cryptoid in the xform_data
1343 * structure. In case when crypto error occurred and crypto
1344 * subsystem has reinited the session, it returns new cryptoid
1345 * and EAGAIN error code.
1347 * This function will be called when we got EAGAIN from crypto
1349 * *new is cryptoid that was returned by crypto subsystem in
1351 * *old is the original cryptoid that we stored in xform_data.
1353 * For first failed request *old == sav->tdb_cryptoid, then
1354 * we update sav->tdb_cryptoid and redo crypto_dispatch().
1355 * For next failed request *old != sav->tdb_cryptoid, then
1356 * we store cryptoid from first request into the *new variable
1357 * and crp_sid from this second session will be returned via
1358 * *old pointer, so caller can release second session.
1360 * XXXAE: check this more carefully.
1363 printf("%s: SA(%p) moves cryptoid %p -> %p\n",
1364 __func__, sav, *old, *new));
1365 KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1367 if (sav->tdb_cryptoid != *old) {
1368 /* cryptoid was already updated */
1370 *new = sav->tdb_cryptoid;
1372 SECASVAR_UNLOCK(sav);
1375 sav->tdb_cryptoid = *new;
1376 SECASVAR_UNLOCK(sav);
1381 ipsec_initialized(void)
1384 return (V_def_policy != NULL);
1388 def_policy_init(const void *unused __unused)
1391 V_def_policy = key_newsp();
1392 if (V_def_policy != NULL) {
1393 V_def_policy->policy = IPSEC_POLICY_NONE;
1394 /* Force INPCB SP cache invalidation */
1397 printf("%s: failed to initialize default policy\n", __func__);
1402 def_policy_uninit(const void *unused __unused)
1405 if (V_def_policy != NULL) {
1406 key_freesp(&V_def_policy);
1411 VNET_SYSINIT(def_policy_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
1412 def_policy_init, NULL);
1413 VNET_SYSUNINIT(def_policy_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
1414 def_policy_uninit, NULL);